corporate bankruptcy reorganizations: estimates from a
Post on 26-Oct-2021
3 Views
Preview:
TRANSCRIPT
Corporate Bankruptcy Reorganizations:
Estimates from a Bargaining Model
Hulya K. K. Eraslan1
Department of Finance
The Wharton School
University of Pennsylvania
Tel: 215-898-9424
Fax: 215-898-6200
eraslan@wharton.upenn.edu
This version: May 17, 2007
1I thank two anonymous referees for their thoughtful comments and suggestions.
I also thank Daniel Bussell, John Geweke, John Kareken, Kenneth Klee, Andrew
McLennan, Antonio Merlo, Philip Bond, Per Stromberg for helpful discussions. I
am grateful to Lynn LoPucki many helpful conversations and data support, and to
Tjomme Rusticus for his excellent assistance on data collection. Financial support
from the National Science Foundation, Graduate School Dissertation Fellowship
of the University of Minnesota, and Ford Motor Company Grant from Rodney L.
White Center for Financial Research of the University of Pennsylvania is gratefully
acknowledged.
Abstract
When a firm files for Chapter 11 bankruptcy in the U.S., negotiations take
place among its claimants to decide what to do with the firm and who gets
what. If an agreement cannot be reached, then the firm is likely to be
liquidated. Consequently, the liquidation value of the firm plays a crucial
role in the deal that is struck among the claimants. In this paper, I use a
novel approach to measure the unobserved liquidation value of a firm that
relies on the information contained in the allocations that are agreed upon in
Chapter 11 negotiations. I do so by estimating a game theoretic model that
captures the influence of liquidation value on the equilibrium allocations
using a newly collected data set. I find that the liquidation values are
higher when the industry conditions are more favorable, and the real interest
rates are higher. I use the estimated model to conduct a counterfactual
experiment to quantitatively assess the impact of a mandatory liquidation
on the equilibrium allocations.
1
1 Introduction
When a firm files for Chapter 11 bankruptcy in the U.S., negotiations take
place among its claimants to decide what to do with the firm and who gets
what. If an agreement cannot be reached, then the firm is likely to be
liquidated. Consequently, the liquidation value of the firm plays a crucial
role in the deal that is struck among the claimants. The goal of this paper
is to use this insight to quantify the liquidation value. In order to do so,
I estimate a structural model of multilateral bargaining that captures the
basic elements of the negotiations in Chapter 11 using a newly collected data
set. This novel approach also allows me to conduct policy experiments to
quantitatively assess the effects of various possible reforms to the bankruptcy
law.
There are two reasons why measuring a firm’s liquidation value is impor-
tant. First, it is the building block for many capital structure theories. These
theories predict the leverage ratio, the maturity of the debt, and the number
of creditors to depend on the liquidation value of the firm in important ways
(see, e.g., Williamson (1988), Harris and Raviv (1990), Aghion and Bolton
(1992), Shleifer and Vishny (1992), Hart and Moore (1994), Bolton and
Scharfstein (1996), Diamond (2004)).1 A higher liquidation value allows the
firm to borrow at more attractive terms, from fewer creditors. Measuring
a firm’s liquidation value is thus central to the understanding of its capital
structure decisions.
Second, liquidation value has important policy implications for the de-
bate on the design of bankruptcy laws. Critics of Chapter 11 argue that it
is costly, keeps inefficient firms operating and gives an unfair advantage to1See Harris and Raviv (1991) for a survey of the capital structure literature.
2
bankrupt firms compared to their non-bankrupt competitors. Consequently,
there is a body of literature calling for repeal of Chapter 11 and proposing
the mandatory liquidation of all bankrupt firms (see, e.g., Baird (1986),
Jackson (1986)). On the other hand, there are scholars who take the view
that large distressed corporations cannot be sold intact, and the difference
between the going concern value and the liquidation value can be substantial
(see, e.g. LoPucki and Whitford (1990), Aghion, Hart and Moore (1992),
Shleifer and Vishny (1992)). Consistent with this view, many countries re-
cently revised their bankruptcy laws toward systems that resemble Chapter
11.
Information on the magnitude of liquidation value relative to going con-
cern value would be useful towards resolving this debate. Measuring the
liquidation value of a firm relative to its going concern value is complicated
because one either observes the liquidation value of a firm or its reorganiza-
tion value. Furthermore, whether a firm chooses to reorganize or liquidate
is itself endogenous. One way to deal with this problem is to first specify a
pricing model that describes the reorganization value as a function of various
firm, industry and macroeconomic characteristics, and then compare the im-
plied prices to actual prices received by the firms in liquidating their assets.
Obviously, this approach is sensitive to the pricing model being correctly
specified.2
2Using this approach, Pulvino (1999) and Stromberg (2000) find that liquidation val-
ues are low relative to going concern values. On the other hand, Maksimovic and Phillips
(1998) and Eckbo and Thorburn (2005) find no evidence of a discount in liquidation rela-
tive to reorganization. In addition to these papers which focus on estimating liquidation
value, there a number of studies are concerned with the effect of liquidation values on
financial contracts. Such papers approximate the liquidation values using various firm
specific information. For example, Alderson and Betker (1995) use management’s re-
3
In this paper, I use a novel approach to measure the unobserved liq-
uidation value of a firm that relies on the information contained in the
agreed upon allocations in Chapter 11 negotiations. I do so by estimating
a game theoretic model that captures the influence of liquidation value on
the equilibrium allocations. The model I consider is a complete information
multilateral bargaining game in which secured creditors, unsecured creditors
and equityholders bargain over what to do with the firm (i.e. reorganize or
liquidate) and how to split the surplus.3 My model differs from the existing
literature on bargaining in Chapter 11 (Baird and Picker (1991), Bebchuck
and Chang (1988), and Kordana and Posner (1999)) by explicitly modeling
the negotiations as a multilateral bargaining game with random proposers
over multiple issues (i.e. the fate of the firm and the division of the surplus).4
In any period when an agreement cannot be reached, there is a chance that
the case is converted to Chapter 7 and the firm is liquidated. In that case,
the payoffs of the claimants are given by an exogenous allocation that de-
pends on the liquidation value of the firm. In equilibrium, the allocation
that is agreed upon reflects this break-up value. As a result, by observing
the actual agreements that are reached, I am able to make inferences about
the liquidation value. I estimate the model using a newly collected data set
ported estimates, Benmelech, Garmaise, Moskowitz (2005) measure liquidation value by
zoning flexibility in the context of commercial loan contracts, and Acharya, Sundaram,
John (2005) use specific assets and intangible assets as proxies for liquidation value.3While the management may have superior information about the value of the firm
in small cases, it is often recognized that in a large bankruptcy case, the creditors come
to the bargaining table with a great deal of information. Therefore, I use a complete
information bargaining model.4Kordana and Posner note that it is difficult to model multiparty negotiations, and
because they found no such models appropriate for Chapter 11, they describe the effects
of having multiple creditors in their model somewhat impressionistically.
4
of 128 Chapter 11 bankruptcies.
The estimates indicate that the liquidation value relative to reorganiza-
tion value is on the average 82% of the reorganization value. For the firms
that actually reorganize, the liquidation value is on the average 75% of the
firm value. The variation among the firms is large enough to suggest that
at least for some firms, the cost of mandatory liquidation is low. However,
the cost mainly depends on aggregate and industry conditions, rather than
observable firm specific variables, making it difficult to design a bankruptcy
procedure that discriminates among firms. I use the estimated model to
conduct a counterfactual experiment to quantitatively assess the impact of
a mandatory liquidation on the equilibrium allocations. I show that such
a policy would hurt all the claimants, but unsecured creditors and equity-
holders would lose disproportionately more than the secured creditors.
This paper also makes a contribution to the estimation of noncooperative
bargaining models. Although there is a large literature that incorporates
noncooperative bargaining models in applied frameworks, there are only
few studies that estimate the parameters of the noncooperative bargaining
models. Notable exceptions are Merlo (1997), and Sieg (2000), Diermeier,
Eraslan, and Merlo (2002, 2003), Watanabe (2006). Merlo (1997) estimates
a stochastic bargaining model of government formation in postwar Italy
to explain the duration of the governments. Sieg (2000) uses generalized
method of moments in the context of a finite horizon asymmetric information
bargaining model using data on medical malpractice disputes. Diermeier,
Eraslan, and Merlo (2002, 2003) estimate stochastic bargaining models to
investigate specific institutional features of parliamentary democracy on the
formation and dissolution of coalition governments. Watanabe (2006) esti-
mates a dynamic bargaining model of dispute resolution with learning using
5
micro data on medical malpractice disputes. To my knowledge, this is the
first paper that estimates a multilateral bargaining model in which the ob-
served variables include how the surplus is allocated among the negotiating
parties.
The rest of the paper is organized as follows. In Section 2, I briefly
summarize the U.S. bankruptcy law. Section 3 lays out the model and char-
acterizes the equilibrium. In Section 4 I describe the data, and in Section
5 I derive the likelihood function that is the basis of the estimation. The
results are presented in Section 6. Section 7 presents the results of simula-
tions to estimate the liquidation value and conducts counterfactual policy
experiments.
2 A Brief Overview of Chapter 11
In the U.S. a firm can file for bankruptcy either under Chapter 7 or under
Chapter 11 of the Bankruptcy Code.5 If the firm files for Chapter 7, then it
is liquidated. If the firm files for Chapter 11, then whether the firm is reor-
ganized or liquidated is determined through negotiations among claimants
of the firm. In practice, large corporations only file under Chapter 11, but
a large firm can still end up in Chapter 7 through the conversion of the
case to Chapter 7 by the court. Since liquidation sets the framework for the
bargaining under Chapter 11, I start by describing the liquidation procedure
under Chapter 7.
Chapter 7: When a firm files for Chapter 7, an automatic stay goes into
effect that prevents the creditors seizing the firm’s assets. A trustee is ap-5The Bankruptcy Code is available online at http://www4.law.cornell.edu/uscode/11.
6
pointed to liquidate the firm. He sells the firm’s assets either as a going con-
cern or piecemeal. In either case, the proceeds are distributed to claimants
according to a ranking called the absolute priority rule. Under absolute pri-
ority rule, secured creditors are paid first, unsecured creditors are paid next
and the equityholders are paid last.
Chapter 11: The goal of Chapter 11 is to reorganize the firm when its
reorganization value exceeds its liquidation value. This is decided by the
claimants of the firms which include secured creditors, unsecured creditors
and equityholders through a bargaining process. As in Chapter 7, an auto-
matic stay goes into effect during the bankruptcy.
The Chapter 11 bankruptcy process begins when a petition is filed under
Chapter 11 of the Code which is designed to keep the firm operating and to
protect its assets while a reorganization plan is being negotiated.
The plan needs to specify what to do with the firm (i.e. whether to
reorganize or liquidate) as well as the allocations to its claimants in return
for cancellation of their original claims against the firm. The claimants are
grouped into classes according to the type of claims they hold, e.g., secured
creditors, unsecured creditors, equityholders, etc. A plan is confirmed if it
is feasible and all classes accept it.6
Under the “cram-down” provisions of the law, even if a class votes against
the plan, it can be confirmed so long as that class receives a payment equal
to what it would receive under absolute priority rule. A typical firm in
Chapter 11 has debt capacity below what it had prior to bankruptcy. Con-6Acceptance of a plan by a class requires that the holders of a majority of the number
of claims within the class and two-thirds of the amount of debt owed to that class vote in
favor of the plan.
7
sequently, the allocations to the creditors almost always involve common
stock in addition to cash and debt. Because of this, the application of cram-
down is limited in practice. As a result, most plans are consensual since the
cost of verifying that a class is receiving as much as it would under absolute
priority rule is considered to be high. In fact, it is these considerations that
led Congress to enact Chapter 11 so that the court’s role is not that of valu-
ing the business, but rather supervising the negotiations among parties who
have better information about the firm. However, when the firm is deeply
insolvent, it is relatively straightforward to argue that the equityholders are
not entitled to any payment under absolute priority rule. In such cases, the
equityholders are left out of the negotiations and the fate of the firm as well
as who gets what is decided by the rest of the claimants.
The management has the exclusive power to propose the plan during the
first four months.7 Typically it develops the plan with a co-proponent that
represents a claimant class. If a plan cannot be confirmed, then the parties
continue negotiating. If no progress is made towards agreement, then the
court can decide to convert the case to Chapter 7.8
3 Model
I model the negotiations among secured creditors, unsecured creditors and
equityholders in a corporate bankruptcy reorganization as an infinite horizon
multilateral bargaining game with complete information.7Prior to 2005, this exclusive right to propose a plan was often extended beyond the
first four months by the court. Bankruptcy Abuse Prevention and Consumer Protection
Act of 2005 forbids the extension of debtor’s exclusivity period beyond 18 months.8The court can also dismiss the case from the bankruptcy proceedings altogether, in
which case state collection laws apply. This is relatively rare.
8
3.1 The Game
There are three players: secured creditors s, unsecured creditors u and the
equityholders e. The amount of debt that the firm owes to creditor i = s, u
is given by Di. In each period, a player is randomly selected to offer a
proposal that specifies what to do with the firm (reorganize or liquidate)
and how to split the value among the players.9 I assume the probability
of being selected as the proposer for player i is stationary and is given by
πi ≥ 0, πs+πu+πe = 1. Note that it is possible that πi = 0 for some player,
in which case the negotiation takes place between the other two players.10
The reorganization value of the firm is V R, and the liquidation value of
the firm is denoted by V L = γV R. The set of feasible allocations depends on
whether the firm is being reorganized or liquidated. Thus, the set of feasible
allocations is given by
XR = {(xs, xu, xe) ∈ IR3 : xi ≥ 0, xs + xu + xe ≤ V R}
if the firm is to be reorganized, and by
XL = {(xs, xu, xe) ∈ IR3 : xi ≥ 0, xs + xu + xe ≤ V L}
if the firm is to be liquidated. For any allocation x ∈ XR ∪ XL, xi is the9Note that the players can decide to liquidate the firm without converting the case to
Chapter 7, and offer an allocation that differs from vL.10When the firm is deeply insolvent, equityholders may not be part of the negotiations
since verifying that the equityholders are not entitled to anything under absolute priority
rule is almost costless. In the empirical implementation of the model, I allow for this
possibility. I do so by making π = (πs, πu, πe) a function of the solvency of the firm as
well as relative size of secured and unsecured debt owed to the creditors. In order to keep
the bargaining model focused, I delay specifying functional forms for the parameters until
Section 5.
9
disbursement to player i. The present discounted utility for player i resulting
from receiving xi in period t is βtxi where β ∈ (0, 1).
Whenever agreement is not reached, there is a chance that the case is
converted to Chapter 7 at the beginning of the next period. I denote the
probability of continuing bargaining at any period (conditional on having
reached that period) by q. If the case is converted to Chapter 7, then
there is liquidation and the players receive their liquidation payoffs according
to the absolute priority rule: the proceeds are used to pay the secured
creditors first, unsecured creditors next and equityholders last. That is,
player s receives vLs = min{V L, Ds}, player u receives vLu = min{max{V L−
Ds, 0}, Du}, and player e receives vLe = max{V L − Ds − Du, 0}. Notice
that, for any given V L, the liquidation payoff is strictly increasing in γ for
precisely one of the players, and is constant in γ for the other two players.
The game is played as follows: At date 0, player i is selected as the
proposer with probability πi. The proposer chooses a proposal that specifies
what to do with the firm µ ∈ {R,L} where R denotes reorganization and
L denotes liquidation, together with an allocation in Xµ. The other players
respond by either accepting or rejecting the proposal. If both of the other
players accept the proposal, the game ends and the proposal is implemented.
If an agreement is not reached, then the process moves into date 1 with
probability q. With probability 1 − q, bargaining breaks down, the case is
converted to Chapter 7, and players receive their liquidation allocation vL =
[vLs , vLu , v
Le ]. The bargaining process continues until either an agreement is
reached or the case is converted to Chapter 7.
An outcome of this game is a pair (µ, x) where µ ∈ {R,L} denotes the
fate of the firm (i.e. reorganization or liquidation), and x ∈ X µ is the agreed
upon allocation. A history is a sequence of realized proposers and actions
10
taken up to that point. A strategy for player i specifies his intended action
(what to propose and how to vote) as a function of the history. A strategy
profile is a tuple of strategies, one for each player. A strategy profile is
subgame perfect (SP) if, at every history, it is a best response to itself. An
SP outcome and payoff are the outcome and payoff functions generated by an
SP strategy profile. A strategy profile is stationary if the actions prescribed
at any history depend only on the current level of surplus, proposer and offer.
A stationary subgame perfect (SSP) outcome and payoff are the outcome
and payoff generated by an SSP strategy profile.11
Several remarks are in order to motivate the assumptions I make in spec-
ifying the set of players. First, there is no judge in the model. There are
several ways the bankruptcy court can influence the outcome of bargaining:
appointment of trustee, extension or lifting of the exclusivity period, set-
ting deadlines, etc. However, in practice, despite the powers they have, the
bankruptcy judges play insignificant roles in large bankruptcy reorganiza-
tions which represents the focus of this study.12
Second, I do not model the management as a separate negotiating party.
The goal of the negotiations in Chapter 11 is to divide the value of the
firm among the creditors and equityholders (along with deciding what to11It is well know that in multilateral bargaining games like the one considered here, if
players are sufficiently patient, then any allocation of the surplus can be sustained as a
subgame perfect equilibrium payoff regardless of the agreement rule (see, e.g., Baron and
Ferejohn (1989) and Merlo and Wilson (1995)). Stationarity, on the other hand, typically
selects a unique equilibrium (see, e.g., Baron and Ferejohn (1989), Merlo and Wilson
(1995) and Eraslan (2002)). Like these papers, I restrict attention to stationary subgame
perfect equilibria.12Based on interviews with bankruptcy attorneys, LoPucki and Whitford (1993) con-
clude that judicial restraint seems to be the norm in large bankruptcies.
11
do with the firm) in return for cancellation of their pre-bankruptcy claims
against the firm. Unless the management has stock holdings of its own, the
only way it can obtain an allocation for itself is through the design of an
excessive compensation package. However, as LoPucki and Whitford (1993)
point out, the creditors and equityholders are able to control management
excesses through the legal process.13
Alternatively, one might believe that management is aligned with the
equityholders and negotiate solely on their behalf. This is not necessarily
the case because managers of insolvent firms owe a fiduciary duty to both
the creditors and the equityholders. In addition, management turnover is
quite large during reorganizations, and a significant number of the changes
are initiated by the creditors14 which may result in hiring of CEOs that are
more aligned with the creditors. One possibility is to extend the model to
explain not only the division of the surplus among the claimants, but also
to explain an additional observable characteristics (such as management
turnover). It may then be possible to identify whether the management
will side with the creditors or the equityholders using the additional data.
Due to lack of data on additional observable characteristics, I leave such an
extension for future work.
3.2 Characterization of Stationary Subgame Perfect Equi-
libria
In this section, I characterize the set of SSP payoffs. Let vi denote the
SSP payoff of player i. Conditional on no breakdown, the subgame starting13In addition, Gilson and Vetsuypens (1993) show that about one third of the CEOs
are replaced and those who stay experience large reduction in their salaries and bonuses.14See e.g. Gilson (1989), Gilson (1990), LoPucki and Whitford (1993), Betker (1995).
12
at any date is identical to the game starting at date 0. This implies that
vi is equal to the expected payoff of player i from the subgame starting
next period discounted back to the current period. I will also refer to vi
as the continuation payoff of player i. Let xκi denote the share of the firm
value awarded to player i under the equilibrium proposal when the proposer
is κ. Since vi is the continuation payoff of player i, it must satisfy vi =
βq∑κ πκx
κi + β(1 − q)vLi . In order to characterize the equilibrium payoffs,
I need to describe the equilibrium strategies which determine xκi .
First note that the proposer will propose to liquidate the firm if V L > V R
and to reorganize the firm if V L ≤ V R since it results in a larger cake to
be divided among the claimants.15 Thus µ = L if V L > V R and µ = R
if V L ≤ V R, and the value of the firm under the chosen alternative is
V = max{V R, V L}. Now consider an SSP response to a proposal x ∈ Xµ.
Player i votes for the proposal if xi ≥ vi and votes against it if xi < vi.16
Thus, to induce acceptance by player i, a payoff maximizing proposer κ
must offer player i a share equal to xκi = vi, and keep xκκ = V −∑i 6=κ vi for
himself. It follows that
vi = βq
πiV −∑
j 6=ivj
+ (1− πi)vi
+ β(1− q)vLi (1)
for all i. Note that a deviation by player i from the strategies above at15The proposer is indifferent between reorganizing and liquidating when V R = V L. For
conciseness, I assume that the proposer proposes to reorganize when indifferent, but this
assumption does not affect the equilibrium payoffs.16Note that when indifferent, player i votes for the proposal. Otherwise, the proposer
can offer vi + ε where ε is arbitrarily small. For the proposer’s problem to have a solution,
it must be the case that player i accepts when indifferent. Also, note that with at least
3 players, there is another equilibrium in which all proposals are rejected. I ignore this
trivial equilibrium.
13
a single date does not affect the continuation payoffs. As a result, when
a player is the proposer, he cannot gain by offering a smaller share to at
least one of the other players because it would be rejected. Obviously he is
not better off by offering a larger share either. Similarly the non-proposing
players cannot gain by deviating from the strategies specified above at a
single date, and conforming to it thereafter. Thus by the one stage deviation
principle for infinite horizon games (see Theorem 4.2 in Fudenberg and Tirole
(1991)), vi satisfies equation (1) if and only if it is an SSP equilibrium payoff.
This completes the characterization of the equilibrium payoffs.
For this game, it is possible to explicitly solve for the payoffs. Summing
(1) over all i, I obtain∑i vi = β
(qV + (1− q)V L
). Substituting back in
(1) for∑j 6=i vj , I obtain
vi = βq(πiG+ vi) + β(1− q)vLi
where G = V − β(qV + (1− q)V L
)= V (1− β (q + (1− q) min{γ, 1})) is
the gain from proposing. Rearranging yields
vi =β(qπiG+ (1− q)vLi )
1− βq. (2)
In equilibrium, the proposer κ receives G + vκ and player i 6= κ receives
vi. Notice that the allocation to all players depend on the same set of pa-
rameters. In other words, a single parameter vector generates an internally
consistent set of equations that describe the allocation vector that would be
agreed upon in equilibrium.
Since the focus of this paper is on estimating the liquidation value pa-
rameter γ, it is useful to briefly discuss its effect on the equilibrium payoffs.
As mentioned earlier, for any given V L, the liquidation payoff of player i
given by vLi is strictly increasing in γ for precisely one of the players, and
14
is constant in γ for the other players. Let i∗(γ) denote this player. When
γ is around 1, typically i∗(γ) = u since most firms are insolvent but their
reorganization value exceeds the secured claims against them.
If γ > 1, then the proposer’s gain G = V L(1 − β) = γV R(1 − β) is
strictly increasing in γ. Consequently, since vLi is also weakly increasing in
γ, all players are strictly better off as γ increases. Suppose now γ < 1 so
that the reorganization value exceeds the liquidation value. In this case, the
proposer’s gain G = V R(1 − β(q + (1 − q)γ)) is strictly decreasing in γ. If
i 6= i∗(γ), then vLi is constant in γ, and hence an increase in liquidation
value lowers player i’s expected payoff vi. For player i∗(γ), there are two
opposing effects of increasing γ when γ < 1. On the one hand, an increase in
γ decreases the gain from proposing whenever he is proposer. On the other
hand his liquidation payoff is strictly higher. It is straightforward to verify
that the latter effect dominates, and thus player i∗(γ) is always better off
as γ increases.
4 Data
The data I use contains information on 128 large, publicly held firms that
filed for Chapter 11 between 1990 and 1997, and had a confirmed plan by the
end of 2000. The initial sample of companies is obtained from Lynn LoP-
ucki’s database available at www.lopucki.com. This database includes all
Chapter 11 bankruptcies with assets of at least $100 million in 1980 dollars
at the time of bankruptcy filing, and had at least one security registered with
the Securities and Exchange Commission (SEC). For each firm in the sample,
I obtained the case number and information on the bankruptcy court that
handled the case from PACER (Public Access to Court Electronic Records),
15
and tracked the disposition of the case (whether it is confirmed, dismissed or
converted) from LEXIS-NEXIS, BankruptcyData.com and Factiva. Once a
case is confirmed, I obtained the confirmation order from the corresponding
bankruptcy court if the case was still in the court.17 I used the confirmation
order to identify the confirmed plan and the associated disclosure statement.
I then contacted the court again to obtain these documents. For some firms,
I was able to obtain the reorganization plans and/or the disclosure statement
from 10K or 8K filings with SEC, or from the lawyers who were involved with
the case. Additionally Lynn LoPucki kindly shared his collection of these
documents. The firms are then matched with corresponding price data from
CRSP database, financial data from Compustat database as well as with
their 10K filings.
From an initial sample of 183 firms, a subsample was selected to meet
some additional criteria. First, I excluded the cases without a confirmed
plan since my model has predictions only with respect to confirmed cases.
There were 16 such cases. Of these 2 were dismissed, 11 were converted
to Chapter 7, and the disposition could not be determined for 3. Second,
12 bankruptcies were excluded because their reorganization plans and the
disclosure statements were missing. Third, 13 were eliminated because they
were not in the Compustat database. Another 14 were excluded because
they lacked financial data within 2 years prior to the bankruptcy filing.
Altogether these criteria resulted in a sample of 128 corporations that filed
for bankruptcy between 1990 and 1997.
For each firm with a confirmed plan, I collected information on the fate17If the case was archived, I obtained the archive locator information from the court,
and the rest of the document retrieval was from the regional National Archive Facility
that stored the corresponding courts documents.
16
of the firm from the reorganization plans. Of the 128 firms in the sample,
17 liquidated and 111 reorganized under the confirmed plan. For 75 firms,
I had sufficient information to quantify the agreed upon allocations.18 The
payments in a Chapter 11 reorganization are typically made using a mix of
different types of securities (e.g., stocks, bonds) as well as cash. Therefore,
a valuation of these securities is necessary to quantify the accepted alloca-
tions. For stocks, I used the first available prices following emergence from
bankruptcy. The prices were obtained from CRSP, 10Ks and pcquote.com.
Debt is valued at face. Warrants are valued using Black-Scholes formula
(adjusted for dilution effect of warrant exercise). For 35 firms, I was also
able to identify the proposer of the plan. Of these 35 firms, unsecured cred-
itors were the proposer in 29 cases, secured creditors were the proposer in 2
cases and the equityholders were the proposer in the remaining 4 cases.
In order to control for variables that may influence the parameters of the
model, I collected additional information about each firm based on the exist-
ing theoretical and empirical research on the determinants of the liquidation
value. Williamson (1988) stresses the link between the specificity of assets
and their liquidation value. In particular, he argues that firms with more
industry specific assets have lower redeployability, and thus lower liquida-
tion values. Shleifer and Vishny (1992) present a model in which the highest
value users of a firm asset are the other firms in the same industry. In their
model, when a firm in financial distress needs to sell its assets, others firms
in the same industry are likely to be experiencing problems as well. This18For some firms, the total claims of one or more creditor classes was missing. When
that is the case, the bargaining model is unable to generate any prediction with respect
to the allocation due to lack of exogenous data. For another group of firms, information
about the price of the securities that were distributed could not be obtained.
17
in turn implies that liquidation values decrease during industry downturns.
To proxy for the asset specificity, I follow the existing literature and use
research and development as a fraction of sales (RESDEV) and intangibles
as a fraction of total assets (INTANG) as control variables.19 Both of these
variables are computed using the last available financial statement in the
last two years before bankruptcy filing. Also following Stromberg (2000),
I classify the assets into three groups depending on specificity: specific as-
sets, nonspecific assets, and intermediary assets, and compute fraction of
specific and nonspecific assets (SPEC and NONSPEC respectively). As in
Stromberg (2000), specific assets include machinery and equipment and non-
specific assets include cash and marketable securities, land and buildings.
Many firms in my sample lacked information to compute SPEC and NON-
SPEC at the firm level, therefore I use the industry median instead. To
account for the overall industry condition, I compute the fraction of the
firms in the industry that had interest coverage less than one during the
year of firm’s bankruptcy filing, and denote this quantity by FRACDIST.
All industry related variables are computed at the four digit SIC code level.
Finally, I control for the overall macroeconomic condition at the time of the
bankruptcy filing using the real interest rate (REALINT). Since recessions
are often preceded by sharp increases in real interest rates, a higher real
interest rate proxies for worse macroeconomic conditions (see, Hall (1999)
for a model in which higher real interest rates result in liquidation and job
destruction). I compute the real interest rate as the difference between the
3 month T-bill rate and the realized inflation rate in the month of firm’s19Intangibles are assets that have no physical existence in themselves, but represent
rights to enjoy some privilege such as client lists, contract rights, copyrights goodwill,
copyrights, formulae, franchises, goodwill, licenses, patents and trademarks.
18
bankruptcy filing.
I also entertain the possibility that the discount factors may differ across
firms. For example, firms that file for prepackaged bankruptcy20 may have
lower discount factors. I let the dummy variable PREPACK take the value
one if the bankruptcy is prepackaged, and zero otherwise. In addition, there
could be differences in the speed with which the courts handle cases. If
that is the case, the discount factors may be lower at slower courts.21 It is
well known that Delaware and Southern District of New York bankruptcy
courts behave differently than other courts, and in particular, there is some
evidence that they tend to be speedier (see, e.g. Ayotte and Skeel (2004)). I
let the dummy variable DELNY take the value one if the firm’s bankruptcy
was filed either in Delaware or Southern District of New York, and zero
otherwise.
Descriptive statistics of all the variables are reported in Table 1 where
FATE is a dummy variable that takes the value one if the firm reorganized,
and zero if it is liquidated; SECPROP is a dummy variable that takes the
value one if the secured creditors were the proposer, and zero otherwise;
UNSECPROP is a dummy variable that takes the value one if the unsecured
creditors were the proposer and zero otherwise; and EQPROP is a dummy
variable that takes the value one if the equityholders were the proposer and
zero otherwise.20In a prepackaged bankruptcy, the reorganization plan is negotiated outside the
bankruptcy. Once the firm files for bankruptcy, it already has an agreed upon plan and
the role of the court is to formally collect the votes and to confirm the plan.21On the other hand, the firms have some ability choose the venue of their bankruptcy.
One might argue that more impatient firms file at speedier courts. Overall this self-
selection combined with the corresponding speed of court would result in roughly constant
discount factors.
19
5 Econometric Specification
Given the vector of parameters (β, γ, π, q), firm value V and creditor claims
D = (Ds, Du), the model described in Section 3 generates a probability
distribution for the following endogenous variables: the identity of the pro-
poser κ, the fate of the firm given by the choice µ (i.e. whether the firm
is reorganized or liquidated), and the accepted proposal xκ = (xκ1 , xκ2 , x
κ3).
Specifically, I have κ = i with probability πi,
µ(β, γ, π, q,D) =
R if γ ≤ 1
L if γ > 1,(3)
xκj (β, γ, π, q,D) =
vj(β, γ, π, q,D) if j 6= κ
G(β, γ, π, q,D) + vj(β, γ, π, q,D) if j = κ,(4)
where
G(β, γ, π, q,D) = V (1− β (q + (1− q) min{γ, 1})) ,
and
vi(β, γ, π, q,D) =β(qπiG(β, γ, π, q,D) + (1− q)vLi (β, γ, π, q,D))
1− βq.
From the perspective of the players, the choice µ and allocation xκ are de-
terministic conditional on κ. I (the econometrician), however, do not know
all the components of the information set of the players. Hence, from the
perspective of the econometrician, µ and xκ are random variables even condi-
tional on κ. Let Z = (V,Ds, Du, RESDEV, INTANG,SPEC,NONSPEC,
FRACDIST,REALINT, PREPACK,DELNY ) denote the set of exoge-
nous variables, and let Fγ(γ|Z) and fγ(γ|Z) respectively denote the distri-
bution and density of the liquidation value conditional on the exogenous
20
variables Z. Thus, from the perspective of the econometrician, the proba-
bility that the firm is reorganized is given by
Pr(µ = R) = Fγ(1|Z) (5)
and the probability that the firm is liquidated is given by
Pr(µ = L) = 1− Fγ(1|Z). (6)
When the firm is so insolvent that it is straightforward to invoke cram-
down by convincing the judge that equityholders would not get anything
under the absolute priority rule. To allow for this possibility, with some
probability δ, equityholders are left out of the game and the negotiations
proceed between the secured and unsecured creditors. Specifically, I assume
that δ = δ(D,V ) where
δ(D,V ) =exp(δ0 + δ1
Ds+DuV )
1 + exp(δ0 + δ1Ds+Du
V ). (7)
Next I need to specify the proposer probabilities. Obviously, the proposer
selection probabilities depend on the number and the identity of the players
in the bargaining. I also allow them to depend on the size of secured and
unsecured debt claims. Let E = 1 denote the event that equityholders are
part of the negotiations, and E = 0 denote the event that equityholders are
left out of the negotiations. As mentioned above, if Di = 0 for i = s, u, then
i is not part of the negotiations. I assume that
πe = πe(E,D, V ) =
0 if E = 0
exp(αe0+αe1DsV
+αe2DuV
)
1+exp(αe0+αe1DsV
+αe2DuV
)if E = 1,
(8)
πs = πs(E,D, V ) =
0 if Ds = 0
1− πe if Ds > 0 and Du = 0
(1− πe)exp(αs0+αs1
DsV
+αs2DuV
)
1+exp(αs0+α1DsV
+αs2DuV
)if Ds > 0 and Du > 0.
(9)
21
Of course, once πe and πs are specified, πu is residually specified by the
identity πu(E,D, V ) = 1− πe(E,D, V )− πs(E,D, V ).
To reduce the influence of outliers, I assume that the allocations are
observed with error. To minimize the number of error terms needed, I
assume that the total firm value (i.e. the reorganization value, if the firm is
reorganized and the liquidation value, if the firm is liquidated) is observed
correctly, but the split of the surplus is observed with error. This assumption
can be justified, for example, by the fact that some individuals may be in
multiple claimant classes.22
Specifically, I assume that, conditional on the true allocation vector (nor-
malized by the firm value) being xκ, I observe the allocation vector (nor-
malized by the firm value) y from a Dirichlet distribution with parameter
Hxκ, and so the density of y is given by
fy(y|xκ, H) =1
B(Hxκ)Πni=1y
Hxκi −1i ,
where n is the dimension of xκ and y (i.e. the number of players), and
B(Hxκ) denotes the normalizing constant for Dirichlet distribution with
parameter Hxκ. That is,
B(Hxκ) =Πni=1Γ(Hxκi )
Γ(∑ni=1Hx
κi ).
Note that
E(yi|xκ, H) = xκi ,
and
V ar(yi|xκ, H) =xκi (1− xκi )H2(H + 1)
,
22This could be the case if vulture investors diversify their portfolio by purchasing claims
in different classes so that their exposure is limited if a particular class does not receive a
favorable treatment. For example, Oaktree Capital, which is one of the biggest players in
the vulture market, uses such a strategy on a systematic basis.
22
so for H large enough the measurement error is small.
I now derive the likelihood function that represents the basis for the
estimation of the structural model. For use below, I let the set K de-
note the set of proposers, which is equal to set of all players when the
identity of the proposer is not observed, and is singleton containing the
identity of the proposer when it is observed. The contribution of each ob-
servation in the sample to the likelihood function is equal to the proba-
bility of observing the vector of endogenous events (K,µ,E, y) conditional
on the exogenous characteristics Z, given the vector of parameters θ =
(αe0, αe1, α
e2, α
s0, α
s1, α
s2, β, δ0, δ1, Fγ , H, q). Given the structure of the model
and the equilibrium characterization, this probability can be written as
Pr(K,µ,E, y|Z; θ) =∑κ∈K
Pr(E|Z; θ) Pr(κ|E,Z; θ) Pr(µ|Z; θ) Pr(y|µ, κ, Z; θ),
(10)
where
Pr(E = 1|Z; θ) = 1− δ(D,V ), Pr(E = 0|Z; θ) = δ(D,V ), (11)
Pr(κ|E,Z; θ) = πκ(E,D, V ) (12)
Pr(µ = R|Z; θ) = Fγ(1|Z; θ), Pr(µ = L) = 1− Fγ(1|Z; θ), (13)
Pr(y|µ = R, κ, Z; θ) =
∫γ≤1 fy(y|xκ(γ))dFγ(γ|Z; θ))
Fγ(1|Z, θ), (14)
Pr(y|µ = L, κ, Z; θ) =
∫γ>1 fy(y|xκ(γ))dFγ(γ|Z; θ)
1− Fγ(1|Z; θ)(15)
where (suppressing its other arguments) xκ(γ) is given by (4). For some
firms, I only observe the fate of the firm µ. The contribution to the likelihood
of these observations is given by Pr(µ|Z; θ). The log-likelihood function is
obtained by summing the logs of (10) over all the elements in the sample
23
for which I have information on the allocation y and the log of Pr(µ|Z; θ)
over all the elements in the sample for which the only information I have
is the fate of the firm µ. The next step consists of choosing a parametric
functional form for Fγ(γ|Z; θ). Since γ ≥ 0 by definition, I assume that γ is
lognormally distributed. In particular, I let
log(γ) = log(γ0 + γ1FRACDIST + γ2REALINT + γ3RESDEV
+γ4INTANG+ γ5SPEC + γ6NONSPEC) + ε, (16)
where ε ∼ N(0, σ2γ). Furthermore, I let
β =exp(β0 + β1PREPACK + β2DELNY )
exp(β0 + β1PREPACK + β2DELNY ) + 1. (17)
6 Results
Table 2 presents the maximum likelihood estimates of the model (αs, αe, β0, β,
δ,Γ, H, σγ , q) where αi = (αi0, αi1, α
i2) for i = s, e, β = (β0, β1, β2), δ =
(δ0, δ1) and Γ = (γ0, . . . , γ6).
An inspection of Table 2 shows that except for the coefficient of nonspe-
cific assets, the sign of the parameters conform to priors. Liquidation value
relative to reorganization value is higher when the industry conditions are
favorable (γ1 < 0), the real interest rates are higher (γ2 > 0), research and
development expenses are lower (γ3 < 0), fraction of intangibles are lower
(γ4 < 0) and fraction of firm specific assets are lower (γ5 < 0). One would
also expect the liquidation value relative to reorganization value to be higher
when the fraction of nonspecific assets is higher. However, the parameter
estimate for γ6 is negative although it is not statistically significant. In fact,
except for the coefficients of industry and macroeconomic conditions, the co-
efficients of the variables describing the distribution of γ are all insignificant.
24
Likewise, the coefficients for the variables that describe the discount factor
are not significant even though they have the expected signs: the discount
factors are higher in bankruptcies that are not prepackaged (β1 < 0) and
filed in Delaware or Southern District of New York (β2 > 0).
As one would expect, the proposer selection probability of the secured
creditors (relative to the proposer selection probability of the unsecured
creditors) depend on the relative composition of the secured and unsecured
debt. In particular, holding δ and πe fixed, the secured creditors are more
likely to be selected as the proposer when the secured claims are higher
(αs1 > 0) and the unsecured claims are lower (αs2 < 0). On the other hand,
the proposer selection probability for the equityholders is lower both when
the secured claims are higher (αe1 < 0) and when the unsecured claims
are higher (αe2 < 0). Furthermore, the parameter estimates for αe1 and
αe2 are not statistically different from each other. This is reasonable: the
equityholders’ bargaining power depends only on the total debt, but not on
how it is composed.
The estimate of β is 0.94 with a standard error of 0.224.23 This implies
that the cost of staying in Chapter 11 is on the average 6% of the firm
value per period. Existing studies that estimate the costs of bankruptcy
distinguish between the direct costs and the indirect costs but do not dis-
tinguish between variable (per period) and fixed costs (see, e.g., Altman
(1984), Weiss (1990), Cutler and Summers (1998), Andrade and Kaplan
(1998)). Direct costs are the payments that the firm needs to make to third
parties as a result of bankruptcy such as lawyers’ fees, investment banking
fees, etc. The studies mentioned above estimate these costs to be 3%-4%23This is obtained by drawing 1000 β0’s, computing β using equation (17) and then
computing the mean and standard deviation of β over all the draws.
25
of the pre-bankruptcy firm value. The indirect costs are the losses due
to decreased productivity, lost customers, cancelled shipments, etc., and is
harder to measure. The studies mentioned above estimate the indirect costs
to be on the average 10-17% of the pre-bankruptcy firm value. One would
expect that the direct costs are incurred largely per period. On the other
hand, it seems likely that most of the indirect costs are incurred at the time
of the bankruptcy news, but do not vary much by the subsequent stay in
bankruptcy.24 As such, the discount factor is better interpreted as a direct
cost. The estimate of the mean bankruptcy cost is rather large compared to
existing estimates of bankruptcy costs.25 It should be noted, however, that
the asymptotic distribution of the discount factor is skewed to the left, so
the mean bankruptcy cost estimate of 6% per period represents an upper
bound.
Equityholders are more likely to be excluded from the negotiations for
less solvent firms (δ1 > 0). The estimates imply that equityholders are
left out of the negotiations, and therefore, receive nothing in 25.33% of the
cases.26 This is precisely what I observe in the data.
More generally, the model fits the data well in other dimensions as well.
To assess the fit of the model, I present Tables 3-7. In each of these tables,24An example of this is the airlines that operate in bankruptcy currently. Many cus-
tomers were aware of the bankruptcy while the news was still fresh, and coverage in
the media was high, but their awareness dissipated over time as the airlines stayed in
bankruptcy.25Existing studies look at a mix of large and small firms, and their estimates are typically
much smaller for larger firms. Also, as mentioned above, the estimated 3% to 4% is an
estimate of the direct costs during the entire duration of bankruptcy which is on the
average 2 years, whereas mine is a per period estimate.26Note that whether the equityholders are part of the negotiations or not (i.e. E = 0
or E = 1) is exogenous and random.
26
I focus on a different dimension of the data, and I compare the predictions
of the model to the empirical distribution. For each dimension of the data,
I use Pearson’s χ2 test,
NJ∑j=1
[f(j)− f(j)]2
f(j)∼ χ2
J−1
to assess the fit of the model, where f(.) denotes the empirical density func-
tion (histogram) of a given endogenous variable, f(.) denotes the maximum
likelihood estimate of the density function of that variable, N is the number
of observations, and J is the number of bins of the histogram.
In Table 3, I compare the density of the fate of the firm predicted by
the model to the empirical density. Table 4 reports evidence on the fit of
the model to the distribution of proposers. In Tables 5, 6 and 7, I compare
the density of the fraction of the firm value received by secured creditors,
unsecured creditors and equityholders respectively.27 As can be seen from
these tables, the model is capable of reproducing the empirical distribution in
each of these dimensions, and χ2 goodness-of-fit tests reported do not reject
the model at conventional significance levels. Furthermore, the predicted
means are almost identical to the ones observed in the data.
7 Liquidation Value and Counterfactuals
Using the estimates, I compute the expected liquidation value relative to re-
organization value. To do this, I draw 5000 samples of parameter values from
the (estimated) asymptotic distribution of the vector of model parameters27Note that the means of the fraction of the firm value received by secured creditors, un-
secured creditors and equityholders need not add up to one because different observations
can have different number of creditors.
27
(based on the estimated variance-covariance matrix), obtain an estimate of
E[γ|Z] for each firm in the sample, and then compute the average across all
observations. The results in an estimate of 0.8246 with a standard error of
0.026. This estimate implies a sizable cost of liquidation (0.1754) although
substantially smaller than Stromberg’s (2000) estimates (ranging from 29%
to 42%) and closer to the lower end of Pulvino’s (1999) estimates (ranging
from 14% to 46%).
Recall that in the model the players agree to liquidate the firm when it
is optimal to do so. This implies that not all firms are reorganized and the
companies that reorganize are expected to have larger liquidation costs. To
evaluate the extent of this selection on the liquidation values, I compute the
conditional expectations of γ (standard errors in parentheses):
E[γ|Z, γ < 1] = 0.7503 (0.0285),
and
E[γ|Z, γ > 1] = 1.1570 (0.0380).
These results indicate that mandatory liquidation of all firms would have
large costs on the average.
I should point out however that the model abstracts from ex ante con-
siderations. One such ex ante consideration is the effect of deviations from
absolute priority rule that is associated with negotiations in Chapter 11 on
the ex ante decisions taken by the equityholders. For example, Bebchuck
(2002) shows that such deviations may aggravate the moral hazard problem
with respect to project choice and may result in distortions in the investment
choice. Since mandatory liquidation ensures that absolute priority rule is
adhered to, its ex ante benefits may outweigh the ex post costs. This would
28
be the case if liquidation costs were small enough. The results indicate that
not all firms have same liquidation costs.
The standard deviation of γ captures the dispersion in the liquidation
costs. The estimates of the unconditional and condition standard deviations
are (standard errors in parentheses):
St.Dev.[γ|Z] = 0.1959 (0.0349)
St.Dev.[γ|Z, γ < 1] = 0.1364 (0.0132)
St.Dev.[γ|Z, γ > 1] = 0.1448 (0.0388).
Even among the firms for which reorganization is optimal (γ < 1), there
is enough variation to suggest that the gains from reorganization are small
for at least for some firms. For such firms, a mandatory liquidation may
be beneficial if the gain in ex ante welfare is sufficiently high. As a result,
there may be efficiency gains if one could design the law to selectively treat
firms differently (such as the treatment of farmers and airlines in the current
bankruptcy law). However, the results also indicate that liquidation costs
are determined mainly by industry and macroeconomic conditions, rather
than firm specific characteristics. More specifically, the liquidation value
relative to reorganization value is higher when the macroeconomic conditions
are less favorable (as in Hall (99)), but the industry conditions are more
favorable (as in Shleifer and Vishny (1992)). Consequently, it is difficult to
redesign the bankruptcy law for selective treatment of firms based on their
expected liquidation costs.
I next use the estimated model to assess the impact of mandatory liqui-
dation on the equilibrium allocation. Table 8 summarizes the result of this
policy experiment. As can be seen from this table, under a mandatory liqui-
dation, the secured creditors’ share of the firm value increases at the expense
29
of both the unsecured creditors and the equityholders. Notice however that
the firm value is lower under liquidation. For the firms for which I have in-
formation on the agreed upon allocations, the total liquidation value is $34.5
billion compared to total reorganization value of $42.8 billion. As a result,
an average of $100 million would be destroyed under mandatory liquidation.
Consequently, even though the secured creditors receive a larger fraction of
the firm, they are worse off under mandatory liquidation. In particular, they
recover 96% of their claims under mandatory liquidation compared to 100%
under the baseline scenario. For the unsecured creditors, the recovery rate
reduces from 62% to 40%. Note, however, that the policy experiment can-
not be used to evaluate the potential benefits of mandatory liquidation. In
particular, if mandatory liquidation provides a relatively speedy resolution
of distress, then its benefits may exceed its costs.
8 Concluding Remarks
In this paper, I have used a novel approach to measure the liquidation
value through the information contained in the agreed upon allocations in
Chapter 11 negotiations. To do so, I developed a multilateral bargaining that
captures the influence of liquidation value on the equilibrium allocations,
and estimated the bargaining model directly. I used the estimated model to
quantify the effect of a mandatory liquidation on the equilibrium allocation.
Throughout the paper, my focus has been exclusively on the split of the
surplus, and abstracted away from the timing of the agreement. A stochastic
bargaining model in which the firm value changes over time can be used
to explain the observed variations in the timing of the agreement. One
could combine the estimation approach developed in this paper with that
30
in Diermeier, Eraslan, Merlo (2002, 2003)) to make use of the information
contained in this variation. I leave this extension for future research.
References
[1] Acharya, Viral V.; Sundaram, Rangarajan K., and John, Kose. “Cross-
Country Variations in Capital Structures: The Role of Bankruptcy
Codes.” Working Paper, 2005.
[2] Aghion, Philippe; Hart, Oliver, and Moore, John. “The Economics of
the Bankruptcy Reform.” Journal of Law, Economics, and Organization,
8:523-546, 1992.
[3] Aghion, Philippe, and Bolton, Patrick. “An ‘Incomplete Contracts’ Ap-
proach to Financial Contracting.” Review of Economic Studies, 59:473-
494, 1992.
[4] Alderson, Michael J., and Betker, Brian L. “Liquidation Costs and Cap-
ital Structure.” Journal of Financial Economics, 39:45-69, 1995.
[5] Altman, Edward I., “A Further Empirical Investigation of the
Bankruptcy Cost Question.” The Journal of Finance, 39: 1067-1089,
1984.
[6] Andrade, Gregor, and Kaplan, Steven N. “How Costly is Financial (Not
Economic) Distress? Evidence for Highly Leveraged Transactions That
Become Distressed.” The Journal of Finance 53:1443 1493, 1998.
[7] Ayotte, Kenneth M., and Skeel, David A. “Why Do Distressed Com-
panies Choose Delaware? An Empirical Analysis of Venue Choice in
Bankruptcy.” Working Paper, 2004.
31
[8] Baird, Douglas G. “The Uneasy Case for Corporate Reorganizations.”
Journal of Legal Studies, 15:127-147, 1986.
[9] Baird, Douglas G., and Picker, Randal C. “A Simple Noncooperative
Bargaining Model of Corporate Reorganizations.” Journal of Legal Stud-
ies, XX: June 1991.
[10] Baron, David P., and Ferejohn, John A. “Bargaining in Legislatures.”
American Political Science Review, 83: 1181-1206, 1989.
[11] Bebchuck, Lucian A. “Ex Ante Costs of Violating Absolute Priority in
Bankruptcy.” The Journal of Finance, 57:445-460, 2002.
[12] Bebchuck, Lucian A., and Chang, Howard F. “Bargaining and the Divi-
sion of Value in Corporate Reorganization.” Journal of Law, Economics,
and Organization, 8:253-279, 1992.
[13] Benmelech, Efraim; Garmaise, Mark and Moskowitz, Tobias. “Do Liq-
uidation Values Affect Financial Contracts? Evidence from Commer-
cial Loan Contracts and Zoning Regulation.” Quarterly Journal of Eco-
nomics, 120: 1121-1154, 2005.
[14] Betker, Brian L. “Management’s Incentives, Equity’s Bargaining Power,
and Deviations from Absolute Priority in Chapter 11 Bankruptcies.”
Journal of Business, 68:161-183, 1995.
[15] Bolton, Patrick, and David Scharfstein. “Optimal Debt Structure and
the Number of Creditors.” Journal of Political Economy, 104: 1-26, 1996.
[16] Cutler, David, Summers, Lawrence, “The Costs of Conflict Resolution
and Financial Distress: Evidence From the Texaco Pennzoil Litigation.”
Rand Journal of Economics, 19:151172, 1988.
32
[17] Diamond, Douglas, “Presidential Address, Committing to Commit:
Short-term Debt When Enforcement Is Costly.” The Journal of Finance,
59: 1447-1479, 2004.
[18] Diermeier, Daniel, Hulya K. K. Eraslan, and Merlo, Antonio. “A Struc-
tural Model of Government Formation.” Econometrica, 71:27-70, 2003.
[19] Diermeier, Daniel, Hulya K. K. Eraslan, and Merlo, Antonio. “Coali-
tion Governments and Comparative Constitutional Design.” European
Economic Review, 46: 893-907, 2002.
[20] Eckbo, Espen B., Thorburn, Karin S. “Bidding in Mandatory
Bankruptcy Auctions: Theory and Evidence.” ECGI Working Paper
Series in Finance, 68/2005, February 2005.
[21] Eraslan, Hulya K. K. “Uniqueness of Stationary Payoffs in the Baron-
Ferejohn Model.” Journal of Economic Theory, 103: 11-30, 2002.
[22] Franks, Julian R., and Torous, Walter N. “An Empirical Investigation
of U.S. Firms in Reorganization.” The Journal of Finance, XLIV:747-
769, July 1989.
[23] Fudenberg, Drew and Tirole, Jean. Game Theory. The MIT Press. 1991.
[24] Gilson, Stuart C. “Management Turnover and Financial Distress.”
Journal of Financial Economics, 25: 241-262, 1989.
[25] Gilson, Stuart C. “Bankruptcy, Boards, Banks and Blockholders.”
Journal of Financial Economics, 26: 355-387, 1990.
33
[26] Gilson, Stuart C., Kose, John, and Lang, Larry H. P. “Troubled Debt
Restructurings: An Empirical Study of Private Reorganization of Firms
in Default.” Journal of Financial Economics, 27: 315-353, 1990.
[27] Gilson, Stuart C., and Vetsuypens, Michael R. “CEO Compensation in
Financially Distressed Firms: An Empirical Analysis.” The Journal of
Finance, 43: 425-458, 1993.
[28] Hall, Robert E. “Aggregate Job Destruction and Inventory Liquida-
tion.” NBER Working Paper 6912, 1999.
[29] Harris, Milton, and Raviv, Artur. “Capital Structure and the Informa-
tional Role of Debt.” The Journal of Finance, 45:321-349, 1990.
[30] Harris, Milton, and Raviv, Artur. “The Theory of Capital Structure.”
The Journal of Finance, 46:297-355, 1991.
[31] Hart, Oliver, and Moore, John. “A Theory of Debt Based on Inalienabil-
ity of Human Capital.” Quarterly Journal of Economics, 109: 841-879,
1994.
[32] Jackson, T. The Logic and Limits of Bankruptcy Law. Mass.: Harvard
University Press, 1986.
[33] Kordana, Kevin A., and Posner, Eric A. “A Positive Theory of Chapter
11.” New York University Law Review, 74:161, April 1999.
[34] LoPucki, Lynn M., and Whitford, William C. “Bargaining Over Eq-
uity’s Share in the Bankruptcy Reorganization of Large, Publicly Held
Companies.” University of Pennsylvania Law Review, 139:125-196, 1990.
34
[35] LoPucki, Lynn M., and Whitford, William C. “Corporate Governance
in the Bankruptcy Reorganization of Large Publicly Held Companies.”
University of Pennsylvania Law Review, 141:669-800, January 1993.
[36] Maksimovic, Vojislav, and Phillips, Gordon. “Asset Efficiency and Re-
allocation Decision of Bankrupt Firms.” The Journal of Finance, 53:
1495-1532, 1998.
[37] Merlo, Antonio. “Bargaining Over Governments in a Stochastic Envi-
ronment.” Journal of Political Economy, 105:101-131, February 1997.
[38] Merlo, Antonio, and Wilson, Charles A. “A Stochastic Model of Sequen-
tial Bargaining with Complete Information.” Econometrica, 63:371-99,
March 1995.
[39] Pulvino, Todd C. “Do Asset Fire Sales Exist? An Empirical Investi-
gation of Commercial Aircraft Transaction.” The Journal of Finance,
53:939-978, 1998.
[40] Shleifer, Andrei, and Vishny, Robert W. “Liquidation Values and Debt
Capacity: A Market Equilibrium Approach.” The Journal of Finance,
47:1343-1365, 1992.
[41] Sieg, Holger. “Estimating a Bargaining Model with Asymmetric In-
formation: Evidence from Medical Malpractice Disputes.” Journal of
Political Economy, 108: 1006-1021, 2000.
[42] Stromberg, Per. “Conflicts of Interest and Market Illiquidity in
Bankruptcy Auctions: Theory and Test.” The Journal of Finance, 55:
2641-2691, 2000.
35
[43] Sutton, John. “Non-cooperative Bargaining Theory: An Introduction.”
Review of Economic Studies, 53: 709-724, 1986.
[44] Weiss, Larry. “Bankruptcy Resolution: Direct Costs and Violation of
Priority of Claims.” Journal of Financial Economics, 27:285-314, 1990.
[45] Williamson, Oliver E. “Corporate Finance and Corporate Governance.”
The Journal of Finance 43:567-592, 1988.
[46] Watanabe, Yasutora. “Learning and Bargaining in Dispute Resolution:
Theory and Evidence from Medical Malpractice Litigation.” Working
Paper, Northwestern University 2006.
36
Table 1
Descriptive Statistics
Mean Standard Deviation Minimum Maximum
FATE 0.13 0.34 0 1
SECPROP 0.06 0.24 0 1
UNSECPROP 0.83 0.38 0 1
EQPROP 0.11 0.32 0 1
Firm value (V) (million $) 565.58 824.17 39.56 5454.14
Allocation to secured (million $) 244.56 441.68 0.00 2893.79 Allocation to unsecured (million $) 271.41 437.65 3.15 3400.00
Allocation to equity (million $) 49.61 243.88 0.00 2054.14
Allocation to secured/Firm value 42.12% 28.99% 0.00% 94.36% Allocation to unsecured/Firm value 51.97% 28.61% 0.67% 100.00%
Allocation to equity/Firm value 5.90% 11.73% 0.00% 73.00%
Secured claims (million $) 252.66 420.18 0.00 2543.88
Unsecured claims (million $) 534.44 662.73 3.15 3384.30
Secured claims/Firm value 45.83% 33.67% 0.00% 132.36%
Unsecured claims/Firm value 122.24% 103.69% 0.67% 613.36%
RESDEV 0.40% 1.49% 0.00% 11.55%
INTANG 8.55% 13.62% 0.00% 59.73%
SPEC 14.64% 13.07% 0.00% 59.27%
NONSPEC 21.62% 16.87% 0.00% 67.47%
FRACDIST 26.85% 17.25% 0.00% 83.33%
REALINT 1.38% 1.10% -0.36% 3.38%
PREPACK 0.11 0.31 0.00 1.00
DELNY 0.46 0.50 0.00 1.00
Table 2
Maximum Likelihood Estimates
Estimate Standard Error
αs0 -2.205 0.902
αs1 4.141 0.988
αs2 -2.459 0.583
αe0 0.917 0.501
αe1 -2.245 0.494
αe2 -1.909 0.390
β0 24.238 14.202
β1 -0.893 0.626
β2 0.657 0.765
δ0 -3.525 0.783
δ1 1.368 0.401
γ0 -0.144 0.080
γ1 -0.340 0.167
γ2 4.832 2.458
γ3 -1.635 2.590
γ4 -0.286 0.191
γ5 -0.110 0.196
γ6 -0.051 0.163
H 93.081 21.325
σγ 0.2210 0.0315
q 0.927 0.245
Log-Likelihood 131.29
Table 3
Density Function of Fate of the Firm
Fate of the firm Data Model
Reorganization 0.867 0.829
Liquidation 0.133 0.172
χ2 test 1.390
Pr(χ2(1) ≥ 1.490) 0.499
Table 4
Density Function of Proposer
Proposer Data Model
Secured creditors 6% 12% Unsecured creditors 83% 74% Equityholders 11% 13%
χ2 test 1.721
Pr(χ2(2) ≥ 1.549) 0.423
Table 5 Density Function of Allocation to Secured
Creditors as a Fraction of Firm Value
Interval Data Model
0%-15% 19% 19% 15%-30% 13% 13% 30%-45% 19% 16% 45%-60% 13% 18% 60%-75% 17% 17% 75%+ 19% 18%
Mean Allocation to Secured Creditors/Firm Value 46% 46%
χ2 test 1.375
Pr(χ2(5) ≥ 1.442) 0.927
Table 6
Density Function of Allocation to Unsecured Creditors as a Fraction of Firm Value
Interval Data Model
0%-15% 9% 13% 15%-30% 20% 15% 30%-45% 13% 18% 45%-60% 13% 15% 60%-75% 19% 15% 75%+ 25% 24%
Mean Allocation to Unsecured Creditors/Firm Value 52% 52%
χ2 test 3.675
Pr(χ2(5) ≥ 3.704) 0.597
Table 7
Density Function of Allocation to Equityholders as a Fraction of Firm Value
Interval Data Model
0%-15% 71% 70% 15%-30% 12% 16% 30%-45% 8% 6% 45%-60% 3% 3% 60%-75% 0% 2% 75%+ 7% 5%
Mean Allocation to Equityholders/Firm Value 6% 5%
χ2 test 3.247
Pr(χ2(5) ≥ 5.034) 0.662
Table 8 Effect of Mandatory Liquidation on the Allocations
Allocations as
Fraction of Firm Value
Baseline Mandatory Liquidation
Secured creditors 46% 58% Unsecured creditors 52% 44%
Equityholders 5% 2%
top related